Ffmpeg Kinetic Captions

CRITICAL GUIDELINES

Windows File Path Requirements

MANDATORY: Always Use Backslashes on Windows for File Paths

When using Edit or Write tools on Windows, you MUST use backslashes (\) in file paths, NOT forward slashes (/).

Documentation Guidelines

NEVER create new documentation files unless explicitly requested by the user.

---

Kinetic Captions Master Guide (2025-2026)

Quick Reference - Kinetic Effects

---

Why Kinetic Captions Matter

• 85% of social video watched without sound
• Word-level animation increases retention by 25-40%
• Scale/bounce effects signal "this word matters NOW"
• Karaoke highlighting guides viewer timing with audio
• Kinetic typography is expected on TikTok/Shorts/Reels

---

Section 1: Karaoke with Word-Grow Effect

The word-grow highlight is the most requested kinetic effect - words scale up when highlighted, then return to normal size.

Basic Word-Grow Karaoke

[Script Info]
ScriptType: v4.00+
PlayResX: 1080
PlayResY: 1920
WrapStyle: 0

[V4+ Styles]
Format: Name, Fontname, Fontsize, PrimaryColour, SecondaryColour, OutlineColour, BackColour, Bold, Italic, Underline, StrikeOut, ScaleX, ScaleY, Spacing, Angle, BorderStyle, Outline, Shadow, Alignment, MarginL, MarginR, MarginV, Encoding
Style: KaraokeGrow,Arial Black,80,&H00FFFFFF,&H0000FFFF,&H00000000,&H40000000,1,0,0,0,100,100,0,0,1,5,0,2,10,10,280,1

[Events]
Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text
; Word grows when karaoke highlight reaches it
Dialogue: 0,0:00:01.00,0:00:05.00,KaraokeGrow,,0,0,0,,{\k80\t(0,200,\fscx115\fscy115)\t(200,400,\fscx100\fscy100)}This {\k60\t(0,150,\fscx115\fscy115)\t(150,300,\fscx100\fscy100)}is {\k100\t(0,250,\fscx120\fscy120)\t(250,500,\fscx100\fscy100)}AMAZING

Understanding the Animation Tags

{\k80\t(0,200,\fscx115\fscy115)\t(200,400,\fscx100\fscy100)}Word

Breaking it down:
- \k80 = Karaoke duration (80 centiseconds = 0.8 seconds)
- \t(0,200,...) = Animation from 0-200ms
- \fscx115\fscy115 = Scale to 115% (grow)
- \t(200,400,...) = Animation from 200-400ms
- \fscx100\fscy100 = Scale back to 100% (shrink)

Critical: Time Units Differ!

---

Advanced Word-Grow Patterns

Pattern 1: Grow + Color Change

Words grow AND change color when highlighted:

[V4+ Styles]
Style: GrowColor,Arial Black,80,&H00FFFFFF,&H0000FFFF,&H00000000,&H40000000,1,0,0,0,100,100,0,0,1,5,0,2,10,10,280,1
;                                 White     Yellow highlight

[Events]
; Karaoke fill (kf) + grow + color transition
Dialogue: 0,0:00:00.00,0:00:03.00,GrowColor,,0,0,0,,{\kf60\t(0,200,\fscx115\fscy115\c&H00FFFF&)\t(200,400,\fscx100\fscy100)}Check {\kf80\t(0,200,\fscx115\fscy115\c&H00FFFF&)\t(200,400,\fscx100\fscy100)}this {\kf100\t(0,250,\fscx120\fscy120\c&H00FFFF&)\t(250,500,\fscx100\fscy100)}out

Pattern 2: Grow + Bounce (Overshoot)

Spring physics overshoot for more dynamic feel:

; Grows to 120%, bounces back to 95%, settles at 100%
Dialogue: 0,0:00:00.00,0:00:03.00,Style,,0,0,0,,{\k80\t(0,120,\fscx120\fscy120)\t(120,220,\fscx95\fscy95)\t(220,350,\fscx100\fscy100)}Bouncy

Pattern 3: Grow + Outline Pulse

Outline expands during highlight:

; Border/outline grows with text
Dialogue: 0,0:00:00.00,0:00:02.00,Style,,0,0,0,,{\k100\t(0,200,\fscx115\fscy115\bord8)\t(200,400,\fscx100\fscy100\bord5)}IMPACT

---

Word-Grow Timing by Platform

TikTok Optimized (Fast)

; Quick 200ms total animation
{\k60\t(0,100,\fscx115\fscy115)\t(100,200,\fscx100\fscy100)}Fast

YouTube Shorts Optimized (Smooth)

; Smooth 350ms total animation
{\k80\t(0,175,\fscx112\fscy112)\t(175,350,\fscx100\fscy100)}Smooth

---

Section 2: CapCut-Style Word Pop Effects

The word pop effect makes each word appear with a bounce, popularized by CapCut.

Basic Word Pop

[V4+ Styles]
Style: WordPop,Arial Black,84,&H00FFFFFF,&H0000FFFF,&H00000000,&H40000000,1,0,0,0,100,100,0,0,1,6,0,2,10,10,280,1

[Events]
; Each word starts at 50% scale, pops to 115%, settles at 100%
Dialogue: 0,0:00:00.00,0:00:00.50,WordPop,,0,0,0,,{\fscx50\fscy50\t(0,100,\fscx115\fscy115)\t(100,200,\fscx100\fscy100)}This
Dialogue: 0,0:00:00.30,0:00:00.80,WordPop,,0,0,0,,{\fscx50\fscy50\t(0,100,\fscx115\fscy115)\t(100,200,\fscx100\fscy100)}is
Dialogue: 0,0:00:00.50,0:00:01.20,WordPop,,0,0,0,,{\c&H0000FFFF&\fscx50\fscy50\t(0,100,\fscx120\fscy120)\t(100,250,\fscx100\fscy100)}VIRAL

Word Pop with Stagger Delay

Words appear sequentially with overlapping timing:

Elastic Word Pop (Multiple Bounces)

; More dramatic with multiple oscillations
; 40% → 130% → 90% → 105% → 100%
Dialogue: 0,0:00:00.00,0:00:01.00,Style,,0,0,0,,{\fscx40\fscy40\t(0,100,\fscx130\fscy130)\t(100,200,\fscx90\fscy90)\t(200,350,\fscx105\fscy105)\t(350,500,\fscx100\fscy100)}ELASTIC

---

Word Pop with Movement

Pop In from Below

; Word moves up while popping
Dialogue: 0,0:00:00.00,0:00:01.00,Style,,0,0,0,,{\move(540,1100,540,960,0,200)\fscx50\fscy50\t(0,100,\fscx115\fscy115)\t(100,200,\fscx100\fscy100)}BounceUp

Pop In from Side

; Word slides in from right while popping
Dialogue: 0,0:00:00.00,0:00:01.00,Style,,0,0,0,,{\move(1200,960,540,960,0,250)\fscx50\fscy50\t(0,120,\fscx110\fscy110)\t(120,250,\fscx100\fscy100)}SlideIn

Pop with Rotation

; Word rotates while appearing
Dialogue: 0,0:00:00.00,0:00:01.00,Style,,0,0,0,,{\frz-15\fscx50\fscy50\t(0,150,\frz0\fscx115\fscy115)\t(150,300,\fscx100\fscy100)}Twist

---

Section 3: Spring Physics Animation

Natural-feeling bounce using spring physics formulas.

Spring Parameter Guide

Damping Ratio Interpretation

• ζ < 0.5: Very bouncy, multiple oscillations
• ζ = 0.5-0.8: Balanced bounce, 1-2 oscillations
• ζ = 1.0: Critical damping (no overshoot)
• ζ > 1.0: Over-damped (sluggish)

Spring Bounce in ASS

Low Damping (Bouncy)

; Multiple oscillations: 120% → 90% → 105% → 98% → 100%
{\fscx80\fscy80\t(0,100,\fscx120\fscy120)\t(100,200,\fscx90\fscy90)\t(200,300,\fscx105\fscy105)\t(300,400,\fscx98\fscy98)\t(400,500,\fscx100\fscy100)}Bouncy

Medium Damping (Balanced)

; Single overshoot: 115% → 95% → 100%
{\fscx80\fscy80\t(0,120,\fscx115\fscy115)\t(120,250,\fscx95\fscy95)\t(250,400,\fscx100\fscy100)}Balanced

High Damping (Quick Settle)

; Minimal overshoot: 110% → 100%
{\fscx80\fscy80\t(0,150,\fscx110\fscy110)\t(150,300,\fscx100\fscy100)}Quick

Position Spring (Vertical Bounce)

; Word bounces to final position
; Starts 100px below, overshoots 20px above, settles
{\pos(540,1060)\t(0,150,\pos(540,940))\t(150,250,\pos(540,980))\t(250,350,\pos(540,960))}Position Bounce

---

Section 4: Shake and Tremor Effects

Impact emphasis with controlled shake animation.

Horizontal Shake

; Shake left-right (±5px) then settle
{\pos(540,960)\t(0,50,\pos(545,960))\t(50,100,\pos(535,960))\t(100,150,\pos(542,960))\t(150,200,\pos(538,960))\t(200,250,\pos(540,960))}SHAKE

Decaying Impact Shake

; Strong shake that decays: 10px → 6px → 3px → 0px
{\pos(540,960)\t(0,50,\pos(550,960))\t(50,100,\pos(530,960))\t(100,160,\pos(546,960))\t(160,220,\pos(534,960))\t(220,280,\pos(543,960))\t(280,350,\pos(540,960))}IMPACT

Shake + Scale Combo

; Word shakes AND scales for maximum impact
{\fscx100\fscy100\pos(540,960)\t(0,50,\fscx120\fscy120\pos(545,960))\t(50,100,\fscx115\fscy115\pos(535,960))\t(100,200,\fscx100\fscy100\pos(540,960))}BOOM

Shake Amplitude Guidelines

Rule: Shake amplitude < 15% of font size for readability.

---

Section 5: Typewriter and Reveal Effects

Character-by-character text reveal.

Character Reveal (Typewriter)

Each character appears sequentially:

; Manual character timing (50ms per character)
Dialogue: 0,0:00:00.00,0:00:00.05,Style,,0,0,0,,H
Dialogue: 0,0:00:00.05,0:00:00.10,Style,,0,0,0,,He
Dialogue: 0,0:00:00.10,0:00:00.15,Style,,0,0,0,,Hel
Dialogue: 0,0:00:00.15,0:00:00.20,Style,,0,0,0,,Hell
Dialogue: 0,0:00:00.20,0:00:01.00,Style,,0,0,0,,Hello

Typewriter Speed Reference

Word-by-Word Reveal

Less tedious than character reveal:

Dialogue: 0,0:00:00.00,0:00:00.30,Style,,0,0,0,,{\fad(100,0)}This
Dialogue: 0,0:00:00.30,0:00:00.60,Style,,0,0,0,,{\fad(100,0)}is
Dialogue: 0,0:00:00.60,0:00:00.90,Style,,0,0,0,,{\fad(100,0)}word
Dialogue: 0,0:00:00.90,0:00:01.20,Style,,0,0,0,,{\fad(100,0)}by
Dialogue: 0,0:00:01.20,0:00:02.00,Style,,0,0,0,,{\fad(100,0)}word

Reveal with Pop

Combine typewriter timing with pop animation:

; Word appears AND pops
Dialogue: 0,0:00:00.00,0:00:00.50,Style,,0,0,0,,{\fscx50\fscy50\t(0,100,\fscx110\fscy110)\t(100,200,\fscx100\fscy100)}This
Dialogue: 0,0:00:00.25,0:00:00.75,Style,,0,0,0,,{\fscx50\fscy50\t(0,100,\fscx110\fscy110)\t(100,200,\fscx100\fscy100)}is
Dialogue: 0,0:00:00.50,0:00:01.00,Style,,0,0,0,,{\fscx50\fscy50\t(0,100,\fscx110\fscy110)\t(100,200,\fscx100\fscy100)}great

---

Section 6: Color Transitions and Multi-Color Karaoke

Progressive Color Fill (Karaoke)

[V4+ Styles]
; Primary = unfilled color, Secondary = filled color
Style: ColorKaraoke,Arial Black,80,&H00FFFFFF,&H0000FFFF,&H00000000,&H40000000,1,0,0,0,100,100,0,0,1,5,0,2,10,10,280,1
;                                   White      Yellow

[Events]
; \kf creates smooth left-to-right color fill
Dialogue: 0,0:00:00.00,0:00:03.00,ColorKaraoke,,0,0,0,,{\kf80}This {\kf60}fills {\kf100}smoothly

Multi-Color Gradient Karaoke

Words transition through multiple colors:

; Yellow → Orange → Red progression
[V4+ Styles]
Style: GradientK,Impact,80,&H0000FFFF,&H000000FF,&H00000000,&H00000000,1,0,0,0,100,100,0,0,1,5,0,2,10,10,280,1

[Events]
; Use \t to add intermediate color during karaoke fill
Dialogue: 0,0:00:00.00,0:00:02.00,GradientK,,0,0,0,,{\kf80\t(0,400,\2c&H0000A5FF&)}Word1 {\kf100\t(0,500,\2c&H0000A5FF&)}Word2
;                                                                  Orange midpoint

Color Flash on Highlight

; Word flashes bright color then returns
Dialogue: 0,0:00:00.00,0:00:02.00,Style,,0,0,0,,{\k80\c&HFFFFFF&\t(0,100,\c&H00FFFF&)\t(100,400,\c&HFFFFFF&)}Flash

---

Section 7: Platform-Specific Caption Styles

TikTok Viral Style

[V4+ Styles]
Style: TikTokViral,Arial Black,88,&H00FFFFFF,&H0000FFFF,&H00000000,&H40000000,1,0,0,0,100,100,0,0,1,6,0,2,10,10,320,1

[Events]
; Fast pop, high energy
Dialogue: 0,0:00:00.00,0:00:00.40,TikTokViral,,0,0,0,,{\fscx50\fscy50\t(0,80,\fscx118\fscy118)\t(80,160,\fscx100\fscy100)}THIS
Dialogue: 0,0:00:00.20,0:00:00.60,TikTokViral,,0,0,0,,{\fscx50\fscy50\t(0,80,\fscx118\fscy118)\t(80,160,\fscx100\fscy100)}IS
Dialogue: 0,0:00:00.40,0:00:01.00,TikTokViral,,0,0,0,,{\c&H00FFFF&\fscx50\fscy50\t(0,100,\fscx125\fscy125)\t(100,200,\fscx100\fscy100)}VIRAL

TikTok Parameters:

• Font size: 84-96px
• Animation: 160-200ms total
• Pop scale: 115-125%
• Word delay: 150-250ms
• Max 3-5 words on screen

YouTube Shorts Professional

[V4+ Styles]
Style: ShortsPro,Montserrat,76,&H00FFFFFF,&H00FFFFFF,&H00333333,&H80000000,1,0,0,0,100,100,0,0,1,4,2,2,10,10,300,1

[Events]
; Smoother, more professional
Dialogue: 0,0:00:00.00,0:00:00.60,ShortsPro,,0,0,0,,{\fscx80\fscy80\t(0,150,\fscx108\fscy108)\t(150,300,\fscx100\fscy100)}This
Dialogue: 0,0:00:00.30,0:00:00.90,ShortsPro,,0,0,0,,{\fscx80\fscy80\t(0,150,\fscx108\fscy108)\t(150,300,\fscx100\fscy100)}is
Dialogue: 0,0:00:00.60,0:00:01.40,ShortsPro,,0,0,0,,{\fscx80\fscy80\t(0,180,\fscx112\fscy112)\t(180,350,\fscx100\fscy100)}professional

YouTube Shorts Parameters:

• Font size: 72-84px
• Animation: 250-400ms total
• Pop scale: 108-115%
• Word delay: 250-350ms
• Max 4-6 words on screen

Instagram Reels Trendy

[V4+ Styles]
Style: ReelsTrend,Impact,82,&H00FFFFFF,&H00FF00FF,&H00000000,&H00000000,1,0,0,0,100,100,2,0,1,5,0,2,10,10,280,1

[Events]
; Stylish with slight rotation
Dialogue: 0,0:00:00.00,0:00:00.50,ReelsTrend,,0,0,0,,{\frz-5\fscx50\fscy50\t(0,120,\frz0\fscx115\fscy115)\t(120,250,\fscx100\fscy100)}Trendy
Dialogue: 0,0:00:00.25,0:00:00.75,ReelsTrend,,0,0,0,,{\frz5\fscx50\fscy50\t(0,120,\frz0\fscx115\fscy115)\t(120,250,\fscx100\fscy100)}Caption
Dialogue: 0,0:00:00.50,0:00:01.20,ReelsTrend,,0,0,0,,{\c&H00FF00FF&\fscx50\fscy50\t(0,150,\fscx120\fscy120)\t(150,300,\fscx100\fscy100)}Style

---

Section 8: Python Script for Kinetic Caption Generation

A full Python kinetic-caption generator that turns word-level transcripts into animated ASS subtitle files (combining all the word-grow, word-pop, spring-bounce, shake, reveal, and karaoke patterns above) lives in references/complete-generator.md. Load that reference when you need a ready-to-run script rather than building per-effect filter graphs by hand.

Pulse/Breathing Effect

# Text pulses 72 ± 8 pixels at ~1.5 Hz
ffmpeg -i input.mp4 \
  -vf "drawtext=text='SUBSCRIBE':fontsize='72+8*sin(t*9.42)':fontcolor=yellow:x=(w-tw)/2:y=h-150:borderw=3:bordercolor=black" \
  output.mp4
# 9.42 rad/s = 1.5 Hz (1.5 * 2π = 9.42)

Bounce Text Entry

# Text bounces from below (spring physics)
ffmpeg -i input.mp4 \
  -vf "drawtext=text='BOUNCE':fontsize=80:fontcolor=white:x=(w-tw)/2:y='(h-th)/2-30*exp(-t*3)*sin(t*15)':borderw=3:bordercolor=black:enable='gte(t,1)'" \
  output.mp4

Scale Pop (Font Size Animation)

# Text pops from small to normal
ffmpeg -i input.mp4 \
  -vf "drawtext=text='POP':fontsize='if(lt(t,0.3),40+40*t/0.3,80)':fontcolor=white:x=(w-tw)/2:y=(h-th)/2:enable='gte(t,0.5)'" \
  output.mp4

Shake on Impact

# Text shakes with decaying amplitude
ffmpeg -i input.mp4 \
  -vf "drawtext=text='IMPACT':fontsize=100:fontcolor=white:x='(w-tw)/2+15*exp(-t*4)*sin(t*50)':y='(h-th)/2+10*exp(-t*4)*cos(t*47)':enable='gte(t,2)'" \
  output.mp4

Sequential Word Appear

# Multiple drawtext filters for word-by-word reveal
ffmpeg -f lavfi -i "color=c=black:s=1080x1920:d=5" \
  -vf "\
    drawtext=text='THIS':fontsize=80:fontcolor=white:x=(w-tw)/2:y=h/2-80:alpha='if(lt(t,0.5),t*2,1)':enable='gte(t,0.5)',\
    drawtext=text='IS':fontsize=80:fontcolor=white:x=(w-tw)/2:y=h/2:alpha='if(lt(t-1,0.5),(t-1)*2,1)':enable='gte(t,1)',\
    drawtext=text='KINETIC':fontsize=80:fontcolor=yellow:x=(w-tw)/2:y=h/2+80:alpha='if(lt(t-1.5,0.5),(t-1.5)*2,1)':enable='gte(t,1.5)'" \
  kinetic_reveal.mp4

---

Section 10: Best Practices and Troubleshooting

Kinetic Caption Best Practices

1. Don't over-animate: Not every word needs maximum effect 2. Emphasize key words: Use larger scale/longer duration for important words 3. Match tempo: Fast speech = fast animation, slow = smooth 4. Platform consistency: Use platform-appropriate styles 5. Test readability: Animation shouldn't hurt comprehension

Emphasis Hierarchy

Common Issues and Solutions

Timing Mismatch

Problem: Animation doesn't match audio timing Solution: Use word-level timestamps from Whisper, not sentence-level

Jittery Animation

Problem: Text appears to stutter Solution: Ensure animation durations are multiples of frame duration

• 30fps: Use 33ms, 67ms, 100ms, 133ms, etc.
• 60fps: Use 17ms, 33ms, 50ms, 67ms, etc.

Text Too Busy

Problem: Animations distract from content Solution: Reduce scale amplitude (115% → 108%), simplify effects

Poor Mobile Readability

Problem: Text hard to read on phones Solution:

• Increase font size (80px minimum for 1080x1920)
• Increase outline/border (4-6px)
• Use safe margins (50px from edges)

---

Sources

This skill was enhanced with research from:

• ASS Override Tags - Aegisub
• FFmpeg Drawtext Animation
• KFX-GUI Karaoke Template Builder
• CapCut Kinetic Typography Guide
• Kinetic Typography Complete Guide 2026
• OpusClip Text Animation Packs
• Spring Physics for UI Animations
• Easing Functions Reference

---

Related Skills

• ffmpeg-karaoke-animated-text - Karaoke fundamentals
• viral-video-animated-captions - CapCut-style captions
• ffmpeg-animation-timing-reference - Timing formulas
• ffmpeg-captions-subtitles - Subtitle basics
• viral-video-platform-specs - Platform requirements

Complete Kinetic Caption Generator (Python + ASS)

A full Python script that takes word-level timestamps and emits Advanced SubStation Alpha (ASS) captions with kinetic animation. SKILL.md keeps the per-effect recipes; this reference holds the assembled generator.

Complete Kinetic Caption Generator

#!/usr/bin/env python3
"""
kinetic_captions.py - Generate kinetic ASS captions from word-level timestamps

Usage: python kinetic_captions.py transcript.json output.ass [style]
Styles: pop, grow, karaoke, bounce, elastic, shake

Input JSON format (Whisper output):
{
  "segments": [
    {
      "words": [
        {"word": "This", "start": 0.0, "end": 0.5},
        {"word": "is", "start": 0.5, "end": 0.8},
        ...
      ]
    }
  ]
}
"""

import json
import sys
from typing import List, Dict, Tuple

# Animation presets
PRESETS = {
    'pop': {
        'start_scale': 50,
        'peak_scale': 115,
        'end_scale': 100,
        'grow_ms': 100,
        'shrink_ms': 100,
        'word_overlap_ms': 50
    },
    'grow': {
        'start_scale': 80,
        'peak_scale': 115,
        'end_scale': 100,
        'grow_ms': 150,
        'shrink_ms': 200,
        'word_overlap_ms': 0
    },
    'karaoke': {
        'use_karaoke': True,
        'peak_scale': 112,
        'grow_ms': 150,
        'shrink_ms': 200
    },
    'bounce': {
        'start_scale': 80,
        'peak_scale': 120,
        'bounce_scale': 95,
        'end_scale': 100,
        'grow_ms': 100,
        'bounce_ms': 100,
        'settle_ms': 100,
        'word_overlap_ms': 50
    },
    'elastic': {
        'start_scale': 40,
        'peak1_scale': 130,
        'valley_scale': 90,
        'peak2_scale': 105,
        'end_scale': 100,
        'phase1_ms': 80,
        'phase2_ms': 80,
        'phase3_ms': 100,
        'phase4_ms': 120,
        'word_overlap_ms': 80
    },
    'shake': {
        'start_scale': 100,
        'peak_scale': 115,
        'shake_amplitude': 8,
        'shake_cycles': 3,
        'shake_duration_ms': 150,
        'grow_ms': 80
    }
}

# Platform presets
PLATFORMS = {
    'tiktok': {
        'font_name': 'Arial Black',
        'font_size': 88,
        'animation_speed': 0.8,
        'max_words': 5
    },
    'youtube': {
        'font_name': 'Montserrat',
        'font_size': 76,
        'animation_speed': 1.2,
        'max_words': 6
    },
    'instagram': {
        'font_name': 'Impact',
        'font_size': 82,
        'animation_speed': 1.0,
        'max_words': 5
    }
}


def format_time_ass(seconds: float) -> str:
    """Convert seconds to ASS timestamp (H:MM:SS.cc)"""
    h = int(seconds // 3600)
    m = int((seconds % 3600) // 60)
    s = seconds % 60
    return f"{h}:{m:02d}:{s:05.2f}"


def generate_pop_effect(word: str, preset: dict) -> str:
    """Generate pop animation tags"""
    s = preset['start_scale']
    p = preset['peak_scale']
    e = preset['end_scale']
    g = preset['grow_ms']
    sh = preset['shrink_ms']

    return f"{{\\fscx{s}\\fscy{s}\\t(0,{g},\\fscx{p}\\fscy{p})\\t({g},{g+sh},\\fscx{e}\\fscy{e})}}{word}"


def generate_grow_effect(word: str, preset: dict) -> str:
    """Generate grow animation tags (for karaoke highlight)"""
    s = preset.get('start_scale', 100)
    p = preset['peak_scale']
    e = preset['end_scale']
    g = preset['grow_ms']
    sh = preset['shrink_ms']

    if preset.get('use_karaoke'):
        # Karaoke duration calculated externally
        return f"{{\\t(0,{g},\\fscx{p}\\fscy{p})\\t({g},{g+sh},\\fscx{e}\\fscy{e})}}"
    else:
        return f"{{\\fscx{s}\\fscy{s}\\t(0,{g},\\fscx{p}\\fscy{p})\\t({g},{g+sh},\\fscx{e}\\fscy{e})}}{word}"


def generate_bounce_effect(word: str, preset: dict) -> str:
    """Generate spring bounce animation tags"""
    s = preset['start_scale']
    p = preset['peak_scale']
    b = preset['bounce_scale']
    e = preset['end_scale']
    g = preset['grow_ms']
    bm = preset['bounce_ms']
    st = preset['settle_ms']

    t1 = g
    t2 = t1 + bm
    t3 = t2 + st

    return f"{{\\fscx{s}\\fscy{s}\\t(0,{t1},\\fscx{p}\\fscy{p})\\t({t1},{t2},\\fscx{b}\\fscy{b})\\t({t2},{t3},\\fscx{e}\\fscy{e})}}{word}"


def generate_elastic_effect(word: str, preset: dict) -> str:
    """Generate elastic overshoot animation tags"""
    s = preset['start_scale']
    p1 = preset['peak1_scale']
    v = preset['valley_scale']
    p2 = preset['peak2_scale']
    e = preset['end_scale']
    t1 = preset['phase1_ms']
    t2 = t1 + preset['phase2_ms']
    t3 = t2 + preset['phase3_ms']
    t4 = t3 + preset['phase4_ms']

    return f"{{\\fscx{s}\\fscy{s}\\t(0,{t1},\\fscx{p1}\\fscy{p1})\\t({t1},{t2},\\fscx{v}\\fscy{v})\\t({t2},{t3},\\fscx{p2}\\fscy{p2})\\t({t3},{t4},\\fscx{e}\\fscy{e})}}{word}"


def generate_shake_effect(word: str, preset: dict) -> str:
    """Generate shake + scale animation"""
    p = preset['peak_scale']
    amp = preset['shake_amplitude']
    cycles = preset['shake_cycles']
    dur = preset['shake_duration_ms']
    g = preset['grow_ms']

    # Calculate shake timing
    cycle_dur = dur // cycles
    shake_tags = ""

    for i in range(cycles):
        t_start = g + (i * cycle_dur)
        t_mid = t_start + (cycle_dur // 2)
        t_end = t_start + cycle_dur
        offset = amp - (i * 2)  # Decay amplitude
        if offset > 0:
            # Note: position shake requires \pos, complex for per-word
            pass

    return f"{{\\fscx100\\fscy100\\t(0,{g},\\fscx{p}\\fscy{p})\\t({g},{g+dur},\\fscx100\\fscy100)}}{word}"


def generate_karaoke_line(words: List[Dict], preset: dict, style_name: str) -> str:
    """Generate karaoke line with grow effect"""
    if not words:
        return ""

    karaoke_text = ""
    for word_data in words:
        word = word_data['word'].strip()
        duration_sec = word_data['end'] - word_data['start']
        duration_cs = int(duration_sec * 100)  # Centiseconds for \k

        grow_tag = generate_grow_effect(word, preset)
        karaoke_text += f"{{\\k{duration_cs}}}{grow_tag}{word} "

    return karaoke_text.strip()


def create_ass_header(platform: str = 'tiktok') -> str:
    """Create ASS file header with styles"""
    plat = PLATFORMS.get(platform, PLATFORMS['tiktok'])

    return f"""[Script Info]
ScriptType: v4.00+
PlayResX: 1080
PlayResY: 1920
WrapStyle: 0
Title: Kinetic Captions

[V4+ Styles]
Format: Name, Fontname, Fontsize, PrimaryColour, SecondaryColour, OutlineColour, BackColour, Bold, Italic, Underline, StrikeOut, ScaleX, ScaleY, Spacing, Angle, BorderStyle, Outline, Shadow, Alignment, MarginL, MarginR, MarginV, Encoding
Style: KineticPop,{plat['font_name']},{plat['font_size']},&H00FFFFFF,&H0000FFFF,&H00000000,&H40000000,1,0,0,0,100,100,0,0,1,5,0,2,10,10,280,1
Style: KineticGrow,{plat['font_name']},{plat['font_size']},&H00FFFFFF,&H0000FFFF,&H00000000,&H40000000,1,0,0,0,100,100,0,0,1,5,0,2,10,10,280,1
Style: KineticKaraoke,{plat['font_name']},{plat['font_size']},&H00FFFFFF,&H0000FFFF,&H00000000,&H40000000,1,0,0,0,100,100,0,0,1,5,0,2,10,10,280,1
Style: KineticBounce,{plat['font_name']},{plat['font_size']},&H00FFFFFF,&H0000FFFF,&H00000000,&H40000000,1,0,0,0,100,100,0,0,1,5,0,2,10,10,280,1
Style: KineticElastic,{plat['font_name']},{plat['font_size']},&H00FFFFFF,&H0000FFFF,&H00000000,&H40000000,1,0,0,0,100,100,0,0,1,6,0,2,10,10,280,1

[Events]
Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text
"""


def generate_kinetic_captions(transcript_path: str, output_path: str,
                              style: str = 'pop', platform: str = 'tiktok'):
    """Generate kinetic ASS captions from Whisper JSON transcript"""

    with open(transcript_path, 'r', encoding='utf-8') as f:
        data = json.load(f)

    preset = PRESETS.get(style, PRESETS['pop'])
    style_name = f"Kinetic{style.title()}"

    header = create_ass_header(platform)
    events = []

    # Extract all words
    all_words = []
    segments = data.get('segments', [])
    for segment in segments:
        words = segment.get('words', [])
        all_words.extend(words)

    if style == 'karaoke':
        # Group words into lines and generate karaoke
        for segment in segments:
            words = segment.get('words', [])
            if not words:
                continue

            start = words[0]['start']
            end = words[-1]['end']
            karaoke_line = generate_karaoke_line(words, preset, style_name)

            events.append(
                f"Dialogue: 0,{format_time_ass(start)},{format_time_ass(end)},{style_name},,0,0,0,,{karaoke_line}"
            )
    else:
        # Word-by-word animation
        effect_func = {
            'pop': generate_pop_effect,
            'grow': generate_grow_effect,
            'bounce': generate_bounce_effect,
            'elastic': generate_elastic_effect,
            'shake': generate_shake_effect
        }.get(style, generate_pop_effect)

        overlap = preset.get('word_overlap_ms', 50) / 1000

        for word_data in all_words:
            word = word_data['word'].strip()
            start = word_data['start']
            end = word_data['end'] + overlap  # Small overlap for smooth display

            effect_text = effect_func(word, preset)
            events.append(
                f"Dialogue: 0,{format_time_ass(start)},{format_time_ass(end)},{style_name},,0,0,0,,{effect_text}"
            )

    # Write output
    with open(output_path, 'w', encoding='utf-8') as f:
        f.write(header)
        f.write('\n'.join(events))

    print(f"Created {output_path} with {len(events)} caption events ({style} style for {platform})")


if __name__ == "__main__":
    if len(sys.argv) < 3:
        print("Usage: python kinetic_captions.py transcript.json output.ass [style] [platform]")
        print("Styles: pop, grow, karaoke, bounce, elastic, shake")
        print("Platforms: tiktok, youtube, instagram")
        sys.exit(1)

    transcript = sys.argv[1]
    output = sys.argv[2]
    style = sys.argv[3] if len(sys.argv) > 3 else 'pop'
    platform = sys.argv[4] if len(sys.argv) > 4 else 'tiktok'

    generate_kinetic_captions(transcript, output, style, platform)

---

Section 9: FFmpeg Drawtext Kinetic Effects

For simple text overlays without ASS files, use FFmpeg drawtext expressions.